From Feedback to Uptake: EFL Students' Perception and Use of ChatGPT Across Writing Stages

• In Event: AI in Higher Education: Balancing Support, Ethics, and Agency

Abstract

This study investigates how Korean EFL university students interact with ChatGPT during a structured academic writing task and how they integrate AI-generated feedback across writing stages. Using a qualitative research design, the study analyzes interaction logs, written essays, and interviews from 13 students. Analysis of 135 interaction turns identified three levels of feedback integration: direct, modified, and none. The results show that integration varied across writing stages, with the highest level occurring during proofreading. Qualitative findings indicate that students generally trusted ChatGPT’s linguistic support for improving fluency, coherence, and lexical choice, while adopting more selective strategies during idea generation, evidence use, and reflection due to concerns about superficiality, credibility, and ethical ambiguity. Many students viewed ChatGPT as a helpful scaffold but sought to maintain authorial authenticity and academic integrity. These findings highlight the importance of stage-sensitive pedagogical guidance and AI literacy to support critical and responsible use of AI in EFL writing contexts

Authors

• Jiyoon Jung, Korea University

• Sewon Joo, Florida State University

• Insook Han, Korea University